Rob Raisch <>
Date: Tue, 13 Jul 1993 16:05:11 -0400 (EDT)
From: Rob Raisch <>
Subject: Re: REGISTRAR
To: Tim Berners-Lee <>
In-reply-to: <9307131303.AA25630@ >
Message-id: <>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Status: RO

On Tue, 13 Jul 1993, Tim Berners-Lee wrote:

> Dave asks how one can specify particular pererences for
> retrieval of an object in the link.  I don't think
> this is information for the link.  In HTTP it is transferred
> atthe time of retrieval from the server, the HTTP headers
> containing information for stating prefernces for variants
> (language, representation, encoding).
> I thinkwork is needed on parameters to hte representations
> like 
>   image/gif
> is all very well, but 
> image/gif  height=200 width=400 colours=8
> provides better information for the server, allowing it to
> select or generate an image more accurately, and save bandwidth.

Tim, this really begs the following question:

	If I am a browser which only displays monochrome GIFs, and the 
	item is only available as a 24bit colour JPEG, whose responsibility
	is it to do the conversion?

	I'd expect that instead of having the server tell the client that
	it can serve it only as:    image/jpeg  colours=24
	and leave the retrieval decision to the browser, (which most times
	is less powerful and feature-ful than the server), 
	would it not be a better decision to have the browser tell the server 
	what  capabilities it had, and let the server make the proper 
	conversions to serve it in an appropriate form?

	The server is more capable of doing the conversion and making
	intellegent decisions regarding what is required to meet the needs
	of the browser.  

	The other way round the server must tell the client what *possible*
	forms the property can be served in, and this quickly becomes 
	unmanageable.  (Consider that GIFs are available with the following
	attributes:  width, heigth, colour(bits), dots-per-inch, aspect)